Blind sensing for hyperspectral surveillance

Abstract

Described is a blind sensing system for hyperspectral surveillance. During operation, hyperspectral data is captured using a hyperspectral camera as mounted on a mobile platform. The system then forms a signal mixture of a plurality of multi-dimensional signals. The multi-dimensional signals being the captured hyperspectral data of a wide area having a background and an object. The plurality of multi-dimensional signals are then demixed using blind source separation, resulting in separated spectra. Finally, the system detects and recognizes a spectral signature of the object in the separated spectra.

What is claimed is: 
     
       1. A blind sensing system for hyperspectral surveillance, the system comprising:
 one or more processors and a memory, the memory being a non-transitory computer-readable medium having executable instructions encoded thereon, such that upon execution of the instructions, the one or more processors perform operations of:
 forming a signal mixture of a plurality of multi-dimensional signals using a subset of dimensions within the multi-dimensional signals, the multi-dimensional signals being hyperspectral data of a wide area having a background and an object and the subset of dimensions being mixing dimensions; 
 demixing the plurality of multi-dimensional signals using blind source separation in dimensions orthogonal from the mixing dimensions, resulting in separated spectra; and 
 detecting and recognizing a spectral signature of the object in the separated spectra in dimensions orthogonal from the mixing dimensions. 
 
 
     
     
       2. The system as set forth in  claim 1 , further comprising an operation of tracking the object. 
     
     
       3. The system as set forth in  claim 1 , wherein the hyperspectral data is captured using a push-broom hyperspectral camera mounted on a mobile platform. 
     
     
       4. The system as set forth in  claim 1 , wherein in forming the signal mixture, hyperspectral data from a plurality of overlapping sliding windows are integrated. 
     
     
       5. The system as set forth in  claim 1 , wherein in timing the signal mixture, hyperspectral data from a plurality of overlapping sliding windows are integrated. and wherein the plurality of overlapping sliding windows are slid in space and time. 
     
     
       6. The system as set forth in  claim 1 , wherein demixing the plurality of multi-dimensional signals is performed using independent component analysis. 
     
     
       7. The system as set forth in  claim 1 , wherein in forming the signal mixture, hyperspectral data from a plurality of overlapping sliding windows are integrated, and wherein the plurality of overlapping sliding windows are formed using nadir or side-looking views from forward motion of a push-broom hyperspectral camera mounted on a mobile platform. 
     
     
       8. The system as set forth in  claim 1 , wherein detecting and recognizing the spectral signature of the object in the separated spectra is performed using a technique selected from a group consisting of nearest neighbor classification, support vector machine classification, neural network classification, and decision tree classification. 
     
     
       9. The system as set forth in  claim 1 , wherein the detected spectral signature of the object provides a cue to cause the system further perform an operation of zooming in on the object with a camera that operates in the mixing dimension. 
     
     
       10. A computer program product for blind sensing hyperspectral surveillance, the computer program product comprising:
 a non-transitory computer-readable medium having executable instructions encoded thereon, such that upon execution of the instructions by one or more processors, the one or more processors perform operations of:
 forming a signal mixture of a plurality of multi-dimensional signals using a subset of dimensions within the multi-dimensional signals, the multi-dimensional signals being hyperspectral data of a wide area having a background and an object and the subset of dimensions being mixing dimensions; 
 demixing the plurality of multi-dimensional signals using blind source separation in dimensions orthogonal from the mixing dimensions, resulting in separated spectra; and 
 detecting and recognizing a spectral signature of the object in the separated spectra in dimensions orthogonal from the mixing dimensions. 
 
 
     
     
       11. computer program product as set forth in  claim 10 , further comprising an operation of tracking the object. 
     
     
       12. The computer program product as set forth in  claim 10 , wherein the hyperspectral data is captured using a push-broom hyperspectral camera mounted on a mobile platform. 
     
     
       13. The computer program product as set forth in  claim 10 , wherein in forming the signal mixture, hyperspectral data from a plurality of overlapping sliding windows are integrated. 
     
     
       14. The computer program product as set forth in  claim 10 , wherein in forming the signal mixture, hyperspectral data from a plurality of overlapping sliding windows are integrated, and wherein the plurality of overlapping sliding windows are slid in space and time. 
     
     
       15. The computer program product as set forth in  Claim 10 , wherein demixing the plurality of multi-dimensional signals is performed using independent component analysis. 
     
     
       16. The computer program product as set forth in  claim 10 , wherein in forming the signal mixture, hyperspectral data from a plurality of overlapping sliding windows are integrated, and wherein the plurality of overlapping sliding windows are formed using nadir or side-looking views from forward motion of a push-broom hyperspectral camera mounted on a mobile platform. 
     
     
       17. The computer program product as set forth in  claim 10 , wherein detecting and recognizing the spectral signature of the object in the separated spectra is performed using a technique selected from a group consisting of nearest neighbor classification, support vector machine classification, neural network classification, and decision tree classification. 
     
     
       18. The computer program product as set forth in  claim 10 , wherein the detected spectral signature of the object provides a cue to cause the one or more processors to further perform an operation of zooming in on the object with a camera that operates in the mixing dimension. 
     
     
       19. A method for blind sensing hyperspectral surveillance, the method comprising an act of:
 causing one or more processors to execute instructions encoded on a non-transitory computer-readable medium, such that upon execution, the one or more processors perform operations of:
 forming a signal mixture of a plurality of multi-dimensional signals using a subset of dimensions within the multi-dimensional signals, the multi-dimensional signals being hyperspectral data of a wide area having a background and an object and the subset of dimensions being mixing dimensions; 
 demixing the plurality of multi-dimensional signals using blind. source separation in dimensions orthogonal from the mixing dimensions, resulting in separated spectra; and 
 detecting and recognizing a spectral signature of the object in the separated spectra in dimensions orthogonal from the mixing dimensions. 
 
 
     
     
       20. The method as set forth in  claim 19 , further comprising an operation of tracking the object. 
     
     
       21. method as set forth in  claim 19 , wherein the hyperspectral data is captured using a push-broom hyperspectral camera mounted on a mobile platform. 
     
     
       22. The method as set forth in  claim 19 , wherein in forming the signal mixture, hyperspectral data from a plurality of overlapping sliding windows are integrated. 
     
     
       23. The method as set forth in  claim 19 , wherein in forming the signal mixture, hyperspectral data from a plurality of overlapping sliding windows are integrated, and wherein the plurality of overlapping sliding windows are slid in space and time. 
     
     
       24. The method as set forth in  claim 19 , wherein demixing the plurality of multi-dimensional signals is performed using independent component analysis. 
     
     
       25. The method as set forth in  claim 19 , wherein in forming the signal mixture, hyperspectral data from a plurality of overlapping sliding windows are integrated, and wherein the plurality of overlapping sliding windows are formed using nadir or side-looking views from forward motion of a push-broom hyperspectral camera mounted on a mobile platform. 
     
     
       26. The method as set forth in  claim 19 , wherein detecting and recognizing the spectral signature of the object in the separated spectra is performed using a technique selected from a group consisting of nearest neighbor classification, support vector machine classification, neural network classification, and decision tree classification. 
     
     
       27. The method as set forth in  claim 19 , wherein the detected spectral signature of the object provides a cue to cause the one or more processors to further perform an operation of zooming in on the object with a camera that operates in the mixing dimension.